The Vision
Others build for datacenters. We build for people. SuperBird pursues elegant efficiency: models so small they run anywhere, so private they never leave your device, and so considered they feel delightful to use.
Intelligence. Liberated.
Magical intelligence. Microscopic footprint. Maximum privacy.
Today, we’re liberating AI from the datacenter. No more locked APIs, no more surveillance — just personal intelligence that runs on your device.
Manifesto
We believe intelligence should be personal—not rented. Small models, on your device, answering to you.
Principles
- Local‑first: inference on your silicon, zero telemetry.
- Efficiency‑by‑design: ternary/quantized paths before scaling.
- Open craft: documented architectures, reproducible results.
- Human taste: usefulness with an aesthetic—quiet, weird, kind.
What We Build
Robin (personal, instant), Cardinal (professional, precise), and Phoenix (multimodal, perceptive)—compact families that share a common ethos: practical speed, small footprints, and trustworthy behavior.
Commitments
- Formats that belong to users: the versioned, portable .nest format.
- Clear licensing and provenance for responsible adoption.
- APIs that stay beautiful even when constraints get ugly.
The SuperBird Difference
- Lightning fast on CPUs you already have.
- Private by architecture, not by promise.
- Impossibly small without feeling compromised.
- Delightfully weird—serious tech with soul.
Personal, 3–8M params, under 10MB. Instant.
- Ultra‑compact language stack
- Latency‑first inference path
- Personal context adapters